Transformer Architecture for Automating Literary Analysis in High School Literature Curricula
Аннотация
Automating literary analysis in high school curricula aims to enhance student engagement and teacher efficiency by leveraging AI for thematic interpretation, device detection, and structural analysis of texts. Such systems can bridge the gap between traditional teaching and modern educational technology. Existing methods, often reliant on generic Natural Language Processing (NLP) models or manual tagging, struggle with capturing nuanced literary devices, aligning with curriculum-specific rubrics, and ensuring interpretability for classroom use. The proposed framework employs a Vision Transformer (ViT) adapted for text, treating passages as token grids to capture long-range dependencies and contextual patterns more effectively. This architecture resolves existing issues by improving semantic feature extraction, device recognition accuracy, and explainability. It is designed for generating structured outputs such as theme summaries, evidence highlights, and essay scaffolds for teacher-guided learning. Findings indicate enhanced accuracy in literary device detection, improved rubric alignment, and increased interpretability, supporting more effective literature instruction.